Cheng-Shong Wu

Cooperative multiagent congestion control for high-speed networks

An adaptive multiagent reinforcement learning method for solving congestion control problems on dynamic high-speed networks is presented. Traditional reactive congestion control selects a source rate in terms of the queue length restricted to a predefined threshold. However, the determination of congestion threshold and sending rate is difficult and inaccurate due to the propagation delay and the dynamic nature of the networks. A simple and robust cooperative multiagent congestion controller (CMCC), which consists of two subsystems: a long-term policy evaluator, expectation-return predictor and a short-term rate selector composed of action-value evaluator and stochastic action selector elements has been proposed to solve the problem. After receiving cooperative reinforcement signals generated by a cooperative fuzzy reward evaluator using game theory, CMCC takes the best action to regulate source flow with the features of high throughput and low packet loss rate. By means of learning procedures, CMCC can learn to take correct actions adaptively under time-varying environments. Simulation results showed that the proposed approach can promote the system utilization and decrease packet losses simultaneously.

Energy Residue Aware (ERA) Clustering Algorithm for Leach-based Wireless Sensor Networks

Many energy efficient routing protocols have been proposed in sensor networks for different scenarios and various applications in literature. One of the efficient way is to group sensors in the neighboring into clusters and send aggregate data by a designated cluster head. However, such design discipline still can not balance the energy consumption of the entire network. As such, in worst cases, some nodes may be soon exhausted. Consequently, it is difficult to provide effective operation and seamless coverage in such a network. In this paper, we propose a novel clustering associating algorithm, known as Energy Residue Aware (ERA) clustering algorithm, to prolong the lifetime of a sensor network by balancing the energy consumption of the entire network. ERA enables each sensor to associate its cluster head to find a path with maximum energy residue sum instead of a path with the minimum energy consumption. Simulations are conducted to evaluate the performance in terms of the network lifetime and energy distribution among nodes. Numerical results show that proposed ERA can maintain a balanced energy consumption distribution among nodes in a sensor network and thus prolong the network lifetime.

Efficiency analyses on routing cache replacement algorithms

Recent research on router architectures focuses on speeding up the time-consuming routing-lookup procedure by elaborate algorithms and data structures to match the ever-increasing wire-speed of fiber links. However, we observed strong temporal locality in the traffic traces collected from two TANet backbone routers. Thus, a routing cache, which is used to reuse previous routing-lookup results, can significantly offload the routing-lookup module. In this paper, we first introduce our temporal locality analysis. Then we investigate the efficiency of several routing cache replacement algorithms, which includes FIFO, LRU, random and our proposed LFU implementation alternative. The simulation results show that our proposed exponentially decayed LFU scheme provides better performance than the other algorithms, especially under small-size caches.

A delay-based transport layer mechanism for fair TCP throughput over 802.11 multihop wireless mesh networks

Cities worldwide have planned and implemented large-scale wireless mesh network (WMN) deployments. These mesh deployments are expected to provide broadband mobile access to the Internet at a low cost to the user. This paper considers the fairness problem affecting nodes in multihop WMNs. Many existing approaches to coping with the fairness problem are unsuitable because they necessitate modifications to the IEEE 802.11 MAC protocol or node queueing operations. Such modifications require a change in hardware/firmware at every node in the WMN. Thus, these approaches are less favorable from the points of view of interoperability and cost. Without modifying the lower layer protocols, this study identifies TCP parameters that impact throughput fairness and proposes an adjustment to these parameters to reduce frame collisions and to improve throughput fairness. Using simple mathematical formulations and ns2 simulations, this study shows that the frame transmissions from each node can be effectively controlled by properly controlling the delayed ACK timer and by using a suitable advertised window. The proposed method, in addition to fairness, requires fewer buffer resources than other methods. Moreover, it is not sensitive to the carrier sense range. It is also simple and easy to deploy. Copyright

Backup VP preplanning strategies for survivable multicast ATM networks

In this paper, we study several backup path building schemes for a multicast tree in self-healing ATM networks. Point-to-point link and path restoration schemes are extended and modified for point-to-multipoint connection in ATM networks. Three spare capacity sharing strategies are discussed. Among these strategies, one is directly from point-to-point restoration schemes and the other two take advantage of the tree structure of point-to-multicast connection to further increase the utilization of spare capacity. By applying different sharing strategies, we develop several link based and path based restoration schemes. These restoration schemes are formulated as combinatorial optimization problems in which the objective functions are the minimization of bandwidth usage and the constraints are required to satisfy the survivability and physical limitations. The backup bandwidth usage, the average restoration time (ART) and the average processing nodes (APN) are the three performance metrics considered in the paper. By evaluating these performance metrics, we conclude that the new proposed link based restoration scheme, Link Scheme 2, which takes advantage of the tree structure of a multicast connection has the best performance.

A REINFORCEMENT LEARNING APPROACH TO CONGESTION CONTROL OF HIGH-SPEED MULTIMEDIA NETWORKS

ABSTRACT A reinforcement learning scheme on congestion control in a high-speed network is presented. Traditional methods for congestion control always monitor the queue length, on which the source rate depends. However, the determination of the congested threshold and sending rate is difficult to couple with each other in these methods. We proposed a simple and robust reinforcement learning congestion controller (RLCC) to solve the problem. The scheme consists of two subsystems: the expectation-return predictor is a long-term policy evaluator and the other is a short-term rate selector, which is composed of action-value evaluator and stochastic action selector elements. RLCC receives reinforcement signals generated by an immediate reward evaluator and takes the best action to control source flow in consideration of high throughput and low cell loss rate. Through on-line learning processes, RLCC can adaptively take more and more correct actions under time-varying environments. Simulation results have shown that the proposed approach can increase system utilization and decrease packet losses simultaneously in comparison with the popular best-effort scheme.

Performance analysis of TCP over wireless link with dedicated buffers and link level error control

To evaluate the behavior of TCP on a path with a wireless link, we build a wireline-wireless network model, and assume the wireless link is the bottleneck which causes packets to be buffered at the wireline-wireless interface buffer. A reliable link layer protocol is assumed on the wireless link which hides the fluctuations of the wireless medium from TCP. We first build the link layer model and the wireless channel Markov chain model, and use the matrix geometric method to evaluate the packet loss probability. We then analyze the long-term steady-state send rate and throughput of a TCP connection.

Characterizing traffic behavior and providing end-to-end service guarantees within ATM networks

We propose a rate-controlled service discipline for supporting B-ISDN services. According to our approach, traffic streams from different connections can be well regulated at the output of each node based on their rate requirements. Moreover, the traffic envelope of a connection inside the network can be effectively characterized. By assuming a leaky-bucket constrained input source, we further prove that the proposed scheme can provide end-to-end delay and jitter bounds for each connection passing through a multi-hop network. Finally, we make a comparison of related works and show the effectiveness of our algorithm.

Centralized Control and Management Architecture Design for PIM-SM Based IP/MPLS Multicast Networks

In this paper, we propose a centralized control and management (CCM) architecture to support PIM-SM multicast protocol in IP/MPLS network. With the integration of IP and MPLS multicast routing mechanism, we justify that CCM framework can improve multicast capabilities such as dynamic group management, control algorithm for multicast routing and resource reservation. Most current schemes for multicast routing assume that multicast routers participate above mentioned functions as well as data forwarding. Since multicast suffers from control overhead by increasing group size, the multicast router must keep large amount of states for point-to-multipoint (p2mp) routing. In additions, the dynamic member join/leave will increase the complexity of control and management for multicast router. By separating data flow and control flow, and by centralizing control and management plane, we design a distinct CCM architecture for PIM-SM based multicast routing in IP/MPLS network. We present the procedural algorithm for the multicast functionality of CCM including call admission, resource reservation, multicast tree calculation and dynamic group management. Compared to source specific and shared multicast routing, the centralized approach can alleviate the burden of multicast routers. To explain the scalability and efficiency for the proposed CCM architecture, we evaluate and discuss the QoS multicast routing for receiver-initiated multicast tree structure.

Hybrid Tree Based Explicit Routed Multicast for QoS Supported IPTV Service

With rapid growth of broadband network deployment and multimedia streaming development, IP multicast networks soon become a delivery mechanism for real-time broadcasting television content. Unlike analog television broadcasting, IPTV is to provide two-way interactive service for viewers to access the selected program channel with high quality multimedia presentation and fast channel surfing capability. However, the traditional IP multicast network can not meet above requirements for IPTV between service provider and end users. To cope with this problem, we propose a hybrid mechanism, which combines the advantages of RP (Rendezvous Point) based shared tree (RPT) and specific source based shortest path tree (SPT) in traditional IP multicast protocol, called hybrid tree based explicit routed multicast (HT-ERM) approach for IPTV service. In particular, thanks to fast-join/leave/switchover operations for IPTV channel switching and RSVP-TE (Reservation Protocol with Traffic Engineering) for explicit routed path setup, HT-ERM can satisfy QoS guarantee. With channel surfing mode, the low-bit-rate quality streams can be accessed by fast-join/leave through aggregated shared multicast tree. In contrast, the high-bit-rate quality stream can be served by fast switch-over to the specific source channel in watching mode. In our simulation, the proposed HT-ERM mechanism can achieve better performance than other QoS based multicast approaches, in terms of channel switching delay, resource utilization, and blocking ratio for IPTV service.